Motion Study
ScreenTrace
ScreenTrace turns a human pose into a dynamic triangulated mesh. Using the 33-point MediaPipe skeleton as vertices, Bowyer-Watson Delaunay triangulation is computed each frame, filling the space between joints with depth-tinted triangles that shift and breathe with the body. The experiment explores the body as data structure — a moving point cloud that generates its own topology in real time.
Four visual modes: Mesh, Edges, Depth, and Aurora — switchable at the bottom of the screen.
The animation simulates a live pose stream; the full interactive version connects an iPhone camera via PeerJS.